Dynamic Spectrum Access in the Time Domain:
Modeling and Exploiting White Space
Stefan Geirhofer and Lang Tong, Cornell University
Brian M. Sadler, United States Army Research Laboratory
IEEE Communications Magazine 2007
Speaker: Chun Hsu 許君
1
Outline

Introduction
Sensing Methods: An Experimental Testbed
Modeling White Space: A Statistical Approach
Deriving Access Schemes: A Practical Example
Conclusion
Comments





2
Introduction

Spectrum scarcity is not a result of heavy usage of the
spectrum; it is merely due to the inefficiency of the static
frequency allocation.
Dynamic spectrum access (DSA) resolves this paradox by
opening frequency bands to secondary users, provided that
interference to the actual licensee is kept insignificant.
The majority of research, so far, has focused on DSA in
the spatial domain.


3
Introduction – Hierarchical Access Model

This article focuses on



applying this concept in the time domain by exploiting idle
periods between bursty transmissions of multi-access
communication channels
addresses WLAN as an example of practical importance.
Hierarchical Access Model


The basic idea is to open licensed spectrum to secondary users
and limit the interference perceived by primary users.
Based on Hierarchical Access Model, they design the
cognitive radio so that both systems are orthogonal.
4
Introduction – Motivating example (1/2)
a) Complex baseband signal of
an 802.11b-WLAN supporting a
Skype conference call
b) enlarged view of two
subsequent packet transmissions
5
Introduction – Motivating example (1/2)

The existence of sufficient white space raises the
questions:




6
How to exploit this resource in practice?
How complicated is a model that provides adequate prediction
performance?
Given a model, how do we apply it to derive practical access
schemes for the secondary system?
Are such schemes amenable to a real-time implementation,
possibly on a battery powered device with processing
limitations?
Sensing Methods – Testbed
VSA, Vector Signal Analyser: do signal measurement and characterization
7
Sensing Methods – Sensing Strategies

The strategies for detecting busy and idle periods can be
classified according to whether the primary user’s
transmission standard is known.


8
Energy-based detection
Feature-based detection
 improve the detection performance by exploiting standard
specifics.
 For the 802.11b, we can find the start of packets by locking onto
the synchronization preamble of the transmitted packets.
 By decoding the LENGTH field within the preamble, we can
find the exact packet duration.
Modeling White Space

Based on the data collected from testbed, they develop a
statistical model that allows us to predict the channel’s
behavior. Two components:


The states of the channel and their transition behavior
How long the system resides in each of the states
No packet
transmission
9
Modeling White Space - Occupancy Durations


The sojourn time in the TRANSMIT state is primarily
affected by the traffic characteristics and the scheduler
employed in the adapter cards and the wireless router.
IDLE state




10
either due to the contention window or a truly free channel
This suggests a mixture distribution.
the contention period shows an almost uniformly distributed
sojourn time
a free state exhibits heavy-tailed behavior that is well
approximated by a generalized Pareto distribution.
Modeling White Space - Goodness-of-fit analysis for
constant payload UDP traffic
Hyper-Erlang provide a viable
approximation to fat-tailed distribution
which leads to the self-similar traffic.
11
Modeling White Space – goodness-of-fit analysis for Nonstationary traffic
Significant
Level α
The Kolmogorov-Smirnov test (KS-test) tries to
determine if two datasets differ significantly.
12
Bluetooth/WLAN Coexistence

The coexistence between Bluetooth and WLAN in the
unlicensed ISM band is of significant practical concern,



because mutual interference can severely limit the performance
of both systems.
ISM band around 2.4GHz
Adaptive Frequency Hopping


13
The sensing and classification of channels according to their
interference.
The adaptation of the hopping sequence such that “bad”
channels are avoided whenever possible.
Bluetooth 1.0-1.1
2.484 CH14(JPN)
2.472 CH13
2.412 CH1
Collisions resulting from random frequency hopping Adapting to
the environment
14
Bluetooth 1.2 and beyond 1.2 - Adaptive Frequency
Hopping
2.484 CH14(JPN)
2.472 CH13
2.412 CH1
Collisions avoided using Adaptive Frequency Hopping
15
Enhanced Hopping Scheme

At the beginning of each slot, the current channel is
sensed.


If the presence of a WLAN packet is detected, no transmission
takes place since this inevitably would lead to a collision.
If no WLAN packet is detected, a collision still can occur if the
WLAN becomes active during the subsequent slot.




16
By initiating a transmission only with a certain probability γ <=1, even
if the channel is sensed idle.
The probability of collision depends on the WLAN traffic
characteristics.
we employ our model to find the probability that conditioned on the
channel having been idle for k slots.
By obtaining this probability, we can design γ such that collisions with
the WLAN occur with a probability smaller than some interference
constraint.
Performance evaluation
17
Conclusion



We addressed the problem of dynamically accessing
spectrum in the time domain by taking advantage of white
space.
The proposed model statistically captures the medium
access of the WLAN but remains tractable enough to be
used for deriving practical access schemes for the
secondary user.
As an example, we have illustrated how our model could
be used to enhance the coexistence between WLAN and
Bluetooth in the unlicensed ISM band.
18
Comments


The Heuristic Scheme is not defined clearly in this paper.
The general scheme should be taken into consideration to
compare with the Heuristic Scheme in the simulation of
collision probability.
19